www.nature.com/scientificreports

OPEN Uncovering the superior corrosion resistance of iron made via ancient Indian iron‑making practice Deepak Dwivedi1, Jitendra P. Mata2, Filomena Salvemini2, Matthew R. Rowles3, Thomas Becker4 & Kateřina Lepková1*

Ancient Indian iron artefacts have always fascinated researchers due to their excellent corrosion resistance, but the scientifc explanation of this feature remains to be elucidated. We have investigated corrosion resistance of iron manufactured according to traditional metallurgical processes by the Indian tribes called ‘Agaria’. Iron samples were recovered from central India (Aamadandh, Korba district, Chhattisgarh). Iron artefacts are investigated using a range of correlative microscopic, spectroscopic, difraction and tomographic techniques to postulate the hidden mechanisms of superlative corrosion resistance. The importance of manufacturing steps, ingredients involved in Agaria’s iron making process, and post-metal treatment using metal-working operation called hot hammering (forging) is highlighted. This study also hypothesizes the probable protective mechanisms of corrosion resistance of iron. Findings are expected to have a broad impact across multiple disciplines such as archaeology, metallurgy and materials science.

Te Indian subcontinent has been famous among metallurgists, conservators and archaeologists for its ancient iron-making technological ­heritage1. Julef et al. (1996) have highlighted the importance of South Asian iron and steel making technology, particularly in peninsular India during the frst millennium AD­ 2. Signifcant metal- lurgical achievements have been made in Indian subcontinent such as the discovery and application of wootz steel (also called “bulad” in Central Asia)3,4. Solid-state ore reduction was one of the important iron-smelting technologies applied in the iron making. Tis process generates agglomerated bloom as the end-product along with ­slag2. Iron-based archaeometallurgical artefacts are important in demonstrating the ancient technological developments that proceeded and dispersed in the ancient ­times5,6. Te Indian blacksmiths were famous glob- ally and the steel (wootz steel—produced from iron bloom manufactured by Indian tribes) produced by them was used in Damascus steelmaking, which was famous for its excellent mechanical properties­ 7. Indian ancient metallurgical and iron making practices have attracted signifcant attention in the recent past but detailed inves- tigations on the structural properties of the tribe’s made iron are yet to be ­performed8. Te ancient Indian iron-making technology has been famous for producing corrosion-resistant iron. Te Delhi iron pillar, which was made around ffh century AD (~ 1600 years old) is famous for its immunity to rust- ing and a prominent example of the Indian iron making ­process9,10. Te Delhi iron pillar is also an important example of tribal tradition (Indian blacksmiths) of iron making in India. Several theories have been postulated regarding corrosion resistance of the Delhi iron pillar. Some of those refer to the inherent nature of the construc- tion material, such as the selection of pure iron, presence of slag particles and slag coatings, surface fnishing using mechanical operation, phosphate flm formation, or the Delhi’s climate­ 11–17. Te exact reason for the superior corrosion resistance phenomenon remains a mystery. It is not possible to examine the iron from the Delhi Iron pillar as it is an archaeologically protected monument. We have examined iron that was manufactured by Indian blacksmiths (Agaria tribes) using an Indian tra- ditional technique. Agaria tribes are regarded as important tribal community responsible for channelizing the traditional iron and steel making technological growth in various Indian states in the central region, such as Madhya Pradesh and Chhattisgarh, Orissa Eastern part of Uttar Pradesh and Bihar from the ancient times­ 18. It is noteworthy that Agaria tribes were not allowed to reside near to the villages, hence they used to perform iron

1Curtin Corrosion Centre, WA School of Mines: Minerals, Energy and Chemical Engineering, Faculty of Science and Engineering, Curtin University, Perth, Australia. 2Australian Centre for Neutron Scattering, Australian Nuclear Science and Technology Organisation (ANSTO), Lucas Heights, NSW 2234, Australia. 3John de Laeter Centre, Curtin University, Perth, Australia. 4School of Molecular and Life Sciences (Chemistry), Curtin Institute for Functional Molecules and Interfaces, Faculty of Science and Engineering, Curtin University, Perth, Australia. *email: [email protected]

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 1 Vol.:(0123456789) www.nature.com/scientificreports/

making practice in a deep forest and adopted this practice as their means of livelihood. Tey were famous for performing special rituals customs (i.e. worshipping their God called Lohasur) prior to the start of iron-making19. India’s status as a colony of the British Empire until 1947, afected the traditional iron-making practices that were banned by the British rulers. Terefore, the technological and historical knowledge of ancient iron-making by Agaria tribes was lost and only a few clues can be obtained from the travel documents written earlier. Tere are still a few old members of the same tribes who have seen iron making practices in their early childhood. Tese are the only remaining sources of the iron making knowledge other than the travel documents. Agaria tribes’ sound- ness and skills in iron making technology made them travel to Japan where one can view their ­crafs19. Igaki et al. (1986) compared the soundness of corrosion resistance of Japanese and Indian ancient iron and found superior corrosion resistance of the Indian iron­ 20. Igaki along with Prakash travelled to Loharpara, Bastar Chhattisgarh, and were able to locate the ‘Mundia’ tribes (diferent tribes from Agarias) and studied their iron-making tech- nology from furnace making and process implementation’s point of view. However, detailed structure–property evaluation of the iron made by these tribes, using their traditional furnaces have not been performed­ 21. In this study, we have applied a range of advanced microscopy, spectroscopy, difraction, and tomography techniques to unveil the hidden corrosion resistance mechanism of iron made by ancient traditional techniques in India. Te reasons for the corrosion resistance of ancient Indian iron is still not clear. Diferent theories such as the presence of phosphorus in iron, presence of inert slag inclusions in iron, intentionally coated slag lay- ers in iron, use of mechanical forming techniques, etc. have been proposed in diferent studies performed on the diferent ancient Indian iron artefacts. However, the corrosion resistance of iron made by tribal traditional techniques has not been examined to date to the best of our knowledge. Materials and methods Material collection. Members of the ‘Agaria’ tribes living in central India (Aamadandh, Korba (District: Korba), Chhattisgarh) donated the iron samples made through ancient Indian metallurgical techniques and using bloom furnace to D.D. for this investigation. A feld survey was also conducted, and iron ore samples and slags collected from the region where the iron was made. However, the furnaces have been destroyed by the weathering action, with only a few remaining signs evident.

Material characterization. In order to investigate the mechanism of superior corrosion resistance of iron made by tribes using ancient Indian metallurgical techniques, multiple characterization tools capable of char- acterizing the morphological and compositional features at various length scales (µm to nm) were used. Mor- phological features of the corrosion product layer formed on the top of the iron are analysed by feld emission gun electron microscope (FESEM). FESEM image acquisition was conducted at 5 kV. Chemical composition of corrosion products was characterized using energy dispersive spectroscopy (EDS) and X-ray photoelectron spectroscopy (XPS). Te EDS acquisition was done at 15 kV, with the EDS coupled to FESEM.

Laboratory‑based surface characterization techniques. Grazing incidence X‑Ray difraction (GIXRD). We have characterized corrosion products using the GIXRD to confrm the crystalline phases pre- sent in the corrosion products. Rietveld analysis was used for evaluating the phase fraction using GIXRD pat- tern. Te penetration depth of the X-ray is limited inside the iron sample. Terefore, X-ray can reveal the cor- rosion flm. However, the entire iron sample could not be analysed using GIXRD. GIXRD data were collected using a PANalytical Empyrean difractometer using Cu Kα radiation. Te incident beam had a parabolic mirror with 2.3° Soller slits, and was fxed at 2° to the specimen surface. Te difracted beam had 2.3° Soller slits, a 0.18° parallel plate collimator, and a point detector with a 0.05 mm receiving slit. Difraction data were taken over the range 10°–100° 2θ in steps of 0.02°. Te difraction data were collected in this fxed-incident beam geometry in order to enhance the contribution from surface layers. Phase identifcation was carried out using the sofware package EVA14 with the ICDD PDF2 database. Crystal structures for the phases were taken from the ICDD PDF 4 + database. Fluorescence, which is considered as an unwanted phenomena in difraction, depends on the type of X-ray radiation source used (Copper (Cu) Kα radiation in our case) and the type of analysed material (iron sample in our case). Iron causes the development of high fuorescence with Cu Kα radiation which leads to the poor signal to noise ratio in difraction pattern. Terefore, we have chosen neutron difraction, used at the Australia’s Nuclear Science and Technology Organisation (ANSTO), Sydney, to analyse iron samples thoroughly because neutrons penetrate deeper inside the iron sample and provide better signal to noise ratio­ 22.

X‑ray photoelectron spectroscopy (XPS). It is also very important to perform chemical characterization of the top layers of corrosion product flm that is exposed to the atmosphere directly. Terefore, XPS was ­used23. Te XPS analysis was exercised using a Kratos Axis Ultra DLD spectrometer (spatial resolution˜5 µm) with an irra- diation source (operated at 225 W) of monochromatic Al Kα (1486.6 eV). Te analysis chamber was kept main- tained at the pressure of 9 × 10–9 Torr until the analysis was completed. C1s spectrum (284.8 eV) was used for calibrating the electron binding energy scale and a survey scan was carried out with the pass energy of 160 eV. CasaXPS sofware was used for data processing.

Field emission scanning electron microscope (FESEM)‑EDS. Te microstructural characterization of corrosion products formed on iron made through traditional ancient iron-making practices by Agaria tribes was con- ducted using a Field emission gun electron microscope (SEM, Zeiss NEON 40EsB).

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 2 Vol:.(1234567890) www.nature.com/scientificreports/

Scanning transmission electron microscope (STEM)‑EDS. Slag compositional analysis was conducted using FEI Titan G2 80–200 TEM/STEM at 200KV and collected high angle annular dark-feld (HAADF) images along with the scanning transmission electron microscopy (STEM).

Synchrotron and neutron based characterization techniques. Neutron difraction. Neutron dif- fraction data were collected at the ANSTO (ECHIDNA—High resolution powder difractometer, Sydney, Aus- tralia) over the range of 4–173° 2θ in steps of 0.5° using Ge monochromators with 0.15° mosaic spread and vertically oriented linear position sensitive detector 3He of 30 cm height (detector radius = 1.29 m)24.

Neutron tomography. Due to the deeper penetration depth of neutrons inside the iron sample, neutron tomog- raphy was used to extract the corrosion product layer formed on the iron as well as characterized pores’ vol- ume inside the iron in three dimensions. Synchrotron tomography is also used for characterizing ancient iron, however, due to the development of the instrumental artefacts, we have selected the application of neutron tomography for three-dimensional characterization of iron made by tribes through ancient technology. Te neutron tomographic analysis of the samples was carried out on ­DINGO25, the ANSTO imaging station (Syd- ney, Australia) fed by a thermal neutron beam coming from the 20 MW OPAL research reactor in Sydney. High resolution acquisition mode with an L/D ratio of 1000 was used and a feld of view 55 × 55 ­mm2 was set to acquire images with a pixel size of 27 μm. A 6LiF/ZnS scintillation screen (thickness of 50 μm) was used. All the projections were acquired in the step of 0.25°over 360° and an exposure time of 50 s each. Te tomographic reconstructions were obtained by using Octopus ­package26 while ­AVIZO27 was used for data visualization, com- piling and quantifcation.

Synchrotron X‑ray tomography. Synchrotron X-ray computed tomography (CT) experiment was done at Hutch 2B of the Imaging and Medical Beamline (IMBL) at the Australian Synchrotron using the ‘Ruby’ detector. It is a custom-designed detector based on a photo-sensitive device coupled by a bright lens to a suitable X-ray sensitive ­scintillator28. PCO edge sensor, which was controlled by the motorized system has been secured from direct and scatter beam radiation by a mirror. Tis is used to view the scintillator plate (placed orthogonally to the direction of the beam). For this experiment, the sensor was equipped with a Nikon Micro-Nikkor 105 mm/f 2.8 macro lens allowing the slide to be used as a zoom control. A terbium-doped gadolinium oxy-sulfde (Gadox, P43) scintil- lator 12 microns thick was used. During the experiment, the system was tuned to produce 2560 × {772}{763} pixels images giving a feld of view of {33.5 × 10.1}{32.8 × 9.8}mm with a pixel size of {13.1}{12.8} microns. Data were collected using monochromatic X-rays of energies {30}{80}{120} keV, in order to ensure reasonable X-ray transmission. Te pieces were positioned vertically and horizontally on the sample stage such that their centre of rotation kept the region of interest within the feld of view of the detector. Each tomographic scan was collected over a 180 range in 0.10 steps, making 1800 views in total, with an exposure time of {0.07}{0.3}{1.5}s per view.

Small‑angle neutron scattering (SANS). SANS measurement was carried out using QUOKKA SANS instrument at the ANSTO (Sydney, Australia)29. Two detector positions (L = 1.3 m and L = 12 m) are used with a Q range of 0.0055715 to 0.714133 / Å. No signifcant neutron activation of the samples was found afer the ­experiment30. SANS data were ftted using Guinier-Porod law and a combination of a Guinier-Porod law and these details can be found elsewhere­ 31,32. Te theory of SANS is provided in the Supplementary section. Results and discussion Te novelty of our study lies in determining the corrosion resistance characteristics of iron made by traditional methods developed by Indian tribes called ‘Agaria’, by using state-of-the-art analytical methods. Te ancient iron was collected from the Agaria tribe members, who no longer continue the iron-making tradition. Te iron sample was manufactured using ancient technology using a typical bloom furnace. Construction of this type of furnace is well described in the article published by Julef et al. (1996)2. Metal products from an ancient bloom furnace are very rare to fnd and pieces of evidence of furnaces have been destroyed due to weathering actions. It has been suggested in the literature that Agaria tribe’s ancient iron-making technology using bloom furnace was in place even before 1200 AD, but the exact date is not ­available33. Te initial step in this study involved collection of the iron samples from tribes (Suppl. Figure 1 (a) and (b)) and iron ores which were used in making ancient irons by tribes from the forest of Aamadandh, Korba, Chhat- tisgarh, India. We have collected and analysed ore samples from the same region (see Suppl. Figure 1 (c)). In the iron making process, these ores were placed inside the bloom furnace. Bowl-shape bloom furnaces were usually built below the ground level by digging a pit (typically of cylindrical shape) with dimensions of about 800 mm in height and 200 mm in diameter. Te shaf of the furnace was constructed below the 600 mm mark. Te construction design of the furnace was completed by a bowl-shaped hearth with a typical diameter of about 240 mm and a depth of about 100 mm. Te bowl-shaped hearth had a hole for tapping out slag (i.e. removing slag from the furnace), which is the waste product of the iron making process. Images of fnal iron products and slags (by Agaria tribe’s traditional iron-making process) collected from the site are shown in Fig. 1 (a). Te placement of the furnace below the ground level was designed to reduce the excess air inside the furnace. It must be highlighted that air blowing was used to maintain the temperature at around 1150 °C. Iron-making was designed to take around 5–6 h/kg iron production and semi-fused mass of sponge iron block came out at the end with a signifcant amount of slag. Te slags that remained in the fnal iron block were removed using forging technique­ 21. Morphological analysis of corrosion product layer of the iron sample (see Fig. 1 (a)) was conducted using feld-emission scanning electron microscope (FESEM). Figure 1 (a) reveals the formation of a thick corrosion

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 3 Vol.:(0123456789) www.nature.com/scientificreports/

Figure 1. (a) FESEM images of (a) the morphology of the corrosion product flm formed on the surface of iron made through ancient Indian metallurgical methods; (b) fake formation in the flm.

product layer on the top of the iron with obvious cracks. It is evinced from Fig. 1a that cracks were in the order of micrometres in size and the majority of the cracks are in the range of ~ 4–5 µm, and a few large cracks were also observed. At the surface areas with a thicker surface flm, less cracks occurred. Flake formation is noticed in the FESEM image (Fig. 1b) which is a common characteristic of atmospheric corrosion. It is evident from earlier studies that fne fakes are the characteristic for atmospheric corrosion process whereas coarse fakes resemble marine ­corrosion34, which is clearly evident from Fig. 1b. Te corrosion product flm is known to be composed of small-size particles of diferent phases, as evident later in this study. We anticipate that the particle-precipitation phenomenon led to formation of the phases through nucleation and growth that are found in corrosion product ­flm34. Nucleation of the initial phase (of corrosion product) is known to assist in the nucleation of another phase by acting as a seed. We believe that phenomenon would have led to the formation of a thick corrosion ­layer35. Cox et al. (1994) suggested the importance of pre- existing oxide layer for nucleation of secondary phase particles or new oxide phases during the formation of passivation layer­ 35. Terefore, we hypothesize that the nucleation and growth of corrosion product layer are also supported by slag phase particles present in the iron. We found that it is difcult to predict the exact phenomenon of growth of corrosion product flm formation in the uncontrolled corrosion experiments (like in this study) because the nucleation and growth is a complex phenomenon and depends on various factors such as substrate crystallographic nature, interface between the substrate and layer, specifc free surface energy, adhesion energy, etc­ 36. Corrosion product flm is found composed of random shape particles as depicted in FESEM analysis (Fig. 2a,b). Mild heterogeneities in morphological features confrmed the less aggressive atmospheric ­corrosion34. Further features of corrosion product flm was characterised using small-angle neutron scattering (SANS) (Fig. 2c). Scattering data were ftted with a shape-independent model called Guinier-Porod ­model31 which revealed the formation of a rough surface with an interlinked structure on the top of the iron, with a Porod exponent of m = 3 (Fig. 2c). Additionally, shape-independent model ftting was performed to investigate the surface features of the corrosion product flm using SANS. Tese results are well-corroborated with the FESEM analysis which exhibited the formation of a rough heterogeneous corrosion product flm. Te compositional analysis of the corrosion product flm is performed by FESEM-energy dispersive spectroscopy (FESEM-EDS) which confrmed the presence of Fe, C, O, Si and Ca (Fig. 3a,b-EDS mapping). Earlier proposed theories discard the utilization of limestone in ancient Indian iron-making furnaces because of the absence of CaO in slag which led to the high P retainment in iron­ 37. On the contrary, the presence of P and Ca is confrmed by the STEM-EDS analysis of collected slags (Suppl. Figure 2). It is worth mentioning that the removal of P from the iron (to slag) depends on various factors, such as the presence of CaO and other basic compounds. Tis theory is well corroborated with the principles of slag thermodynamics and chemistry that suggest higher P removal with CaO than the FeO­ 38. It is noteworthy that the P can lead to the formation of liquid phosphides at the grain boundaries and cause cracking during mechanical working such as ­forging39,40. Terefore, in order to clarify further that whether CaO was added intentionally during the iron making process to remove P from iron, we have characterized the iron ores, which were used in iron making practice and col- lected from the same site, using XRD. Te XRD analysis (Suppl. Figure 3) confrms the presence of hematite, kaolinite, and anatase phases and excludes the presence of Ca in the ores. We speculate that the addition of Ca to the iron was through the clay used for preparing the bottom part of the furnace along with the fne coal dust. It is worth mentioning that a slanting platform made from clay-coated bamboo was used for sliding the charge into the furnace and therefore some Ca addition is also possible from this operation. However, the probability

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 4 Vol:.(1234567890) www.nature.com/scientificreports/

Figure 2. (a) FESEM image of the corrosion product flm found on the top of the iron made by Agaria tribe (b) FESEM image depicting the inhomogeneity in the shape and size of the corrosion phase particles that are found in corrosion product flm (c) represent the small-angle neutron scattering (SANS) scattering data collected from the iron made through ancient Indian metallurgical methods, scattering data is ftted with the Guinier-Porod model.

of introducing Ca into the melt from the bottom portion of the furnace is more likely compared to the slanting platform used for the charge sliding­ 19. Figure 4 and Suppl. Figure 4 elucidate the outwards segregated difusion of slag constituents such as Si and Ca. (Slag constituents are expected to come from the iron ore used in the iron-making by tribes as discussed later with the help of XRD of iron ore analysis, which confrms the presence of kaolinite.) Tis phenomenon led to the localized dissolution of Fe in the corrosion layer. Metal dissolution is considered as anodic reaction whereas oxygen reduction is known as a cathodic reaction (see the reaction (1)). Signifcant depletion of Fe is observed which afrmed the anodic dissolution. Tis depletion of Fe causes the dif- fusion of Si and Ca that were found segregated at few locations. It is also noteworthy that the difusion of Si and Ca from the surface of iron into the corrosion layer (where particularly Fe was found depleted) is attributed to the raising oxygen potential due to the combined efect of formation of FeO layer and electrochemical erosion­ 41. Bufered surface of iron discards the hydrogen reduction as prominent cathodic reaction such as: 3 2Fe + H O + O = 2FeOOH 2 2 2 (1)

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 5 Vol.:(0123456789) www.nature.com/scientificreports/

Figure 3. (a) FESEM images of corrosion product flm formed on the surface of iron made through ancient Indian metallurgical methods (b) EDS mapping of Fe, O, C, Si, Ca without P (dots appeared in P mapping are due to instrument noise).

It is also noteworthy that initially formed lepidocrocite is further transformed to goethite and spinel iron oxide-hydroxide. Tis continuous transformation resulted in the formation of maghemite and hematite (as observed). Hematite is the most stable phase among these two phases due to highly negative Gibbs free ­energy42. FESEM-EDS mapping (see Fig. 3 and Suppl. Figure 4) showed the formation of a slag layer consisting of Ca and Si with Fe and O in the corrosion layer. It was well documented in earlier research that ion-exchange driven non- + 43 stoichiometric removal of cations drives the inward difusion of hydrogen species (H­ 3O , ­H2O) . Atmospheric corrosion can lead to this condition that ofers outward difusion of cations, while alkali ions in this study are either dissolved or difused inwards. We, therefore, speculate that Fe acted as an anodic cell due to the formation of a localized galvanic cell composed of Fe as an anode, whereas slag phases acted as the cathode and appeared to be dissolved at certain areas of the corrosion layer. Tis localized chemical imbalance led to changes in the chemical potential which drove the difusion of Si and Ca and assisted in their segregation­ 43. Te presence of slag phases in the corrosion layer of iron was further confrmed from the EDS spectra from various locations at the surface (see Suppl. Figure 4). Ca, Si, Al were detected along with Fe, C, and O. Localized dissolution of Fe and non-homogenous distribution of slag elements were further noticed from the EDS data, confrming our earlier argument of localized cell formation that leads to difusion of ions. Formation of bubbles in the iron was also noticed, which is a typical characteristic of iron made in bloom furnace­ 44. EDS analysis confrms the presence of Fe, O, Si, Al, Ca, and Ti; without any presence of P. It is speculated that the surface segregation of Ti would have played an important role in making the Agaria tribe’s iron corrosion resistance. A recent study 45 conducted with TiO­ 2-clay composite also confrmed the hydrophobic nature of this ­composite . Terefore, we hypothesize that the surface segregation of Ti would have supported in the formation of the protective passive- flm on iron and helped Agaria tribe’s iron to be corrosion resistant. Te source of Ti addition in iron was iron ores which were used in the iron-making process and contain 4 wt% of ilmenite (see Suppl. Fig. 3). Terefore, it is believed that Ti was not alloyed in iron intentionally. We have carried out STEM-EDS (elemental mapping) (see Suppl. Figure 2) of slag, which is a by-product of the iron-making process. Ti was not evidenced from the STEM-EDS mapping. Hence, it is believed that Ti was added in the iron material, namely from iron ore with anatase that was used in the iron-making process by tribes. Te role of Ti in the prevention of metal corrosion is well ­described46. In order to investigate the interface between the iron and the corrosion flm, cross-sectional FESEM was performed (see Fig. 4a,b) and shows the formation of a coherent interface with a thick corrosion layer. Cross-sectional FESEM images confrmed (within the resolution limit of FESEM) the absence of pores and pits at the interface and showed coherent interface as one of the paramount factors responsible for high corrosion

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 6 Vol:.(1234567890) www.nature.com/scientificreports/

Figure 4. Cross-sectional FESEM images of corrosion product flm formed on iron made through ancient Indian metallurgical methods depicting (a) interface formation between corrosion product flm and substrate (iron) (b) interface formation between corrosion product flm and substrate (iron) at diferent location and in diferent (view) orientation depicting the small holes on flm.

resistance of the ancient iron. It is worth mentioning that defects such as holes were noticed on the top of the corrosion product layers. However, we could not identify from the FESEM data whether these holes extended to the iron surface. We have therefore employed neutron tomography as mentioned later. Correlative microscopic and spectroscopic techniques are found useful for in-depth analysis of corrosion products formed on the ancient iron ­artefacts47,48. In this study, corrosion layer was analyzed using GIXRD. Te formation of hematite was exhibited in the difraction pattern obtained by GIXRD analysis (Suppl. Figure 5). GIXRD analysis of corrosion layer has confrmed the presence of hematite (Fe­ 2O3), quartz (SiO­ 2) and calcite ­(CaCO3), while Rietveld analysis proved the formation of Fe­ 2O3 (70wt%) along with SiO­ 2 (19wt%) and CaCO­ 3 (11wt%) (Suppl. Figure 5). Tese protective corrosion products might have led to high corrosion resistance­ 47. However, the signal to noise ratio was poor and the ancient iron sample was thus further characterized using neutron difraction. In fact, neutrons are electrically neutral and can penetrate deeper inside the ancient iron. Neutron difraction confrmed the presence of iron (Fe), cementite (Fe­ 3C, cohenite in mineralogy) and magh- emite (ϒ-Fe3O4), but a few difraction peaks remained unclassifed. Rietveld analysis of neutron difraction pat- tern confrmed the association of ~ 92 wt% of Fe with 1.1 wt% of Fe­ 3O4 and 1.7 wt% of ­Fe3C (Suppl. Figure 6). Unidentifed phases were about 5 wt%. Te unidentifed peak positions (at 2θ°) were 40.62°, 42.38°, 64.49°, 76.86°, 96.73°, and 115.34°. Neutron difraction data confrmed the formation of BCC α-Fe. Detection of BCC α-Fe made us believe that Agaria’s furnace was suitable to be operated below 1000 °C. Te corrosion product flm was found composed of hematite, maghemite and slag phases as confrmed by GIXRD and neutron difraction. Maghemite is expected to be formed underneath these layers as detected in neutron difraction analysis (Suppl. Figure 6), which is a technique used for bulk-material characterization. Among these iron oxide polymorphs, hematite is the most stable iron oxide with Gibbs free energy of forma- tion of –744.4 ± 1.3 kJ mol-1 whereas maghemite is considered a less stable polymorph with Gibbs free energy of formation of –731.4 ± 2.0 kJ mol-1 at 298 K and 1 bar ­pressure49. It is believed that the formation of oxides could have occurred through internal oxidation ­mechanism50. Internal oxidation allows oxygen to enter into the material and this difusion process leads to the sub-surface precipitation of oxides of alloying elements. Te concentration of these solute elements is imperative and decides the transition between internal to external oxidation. We have noticed segregated Si in EDS mapping, which indicates the long-range difusion of Si from the matrix (parent metal). It is also worth to note that this difusion makes the chemical potential of Si (in the oxide zone) lower than the surrounding matrix and is considered the driving force for the Si difusion. However, in EDS mapping analysis, we have noticed the widespread Si and Ca with O which confrms the difusion of these elements from the grain/sub-grain of the matrix (iron) to the oxygen difusion zone. We have confrmed the formation of several oxide zones (multilayer oxide formation), which is a typical characteristic of internal oxidation, by using various techniques (GIXRD, neutron difraction, XPS, synchrotron and neutron tomog- raphy). Hematite is the most stable phase among the iron ­oxides49 and maghemite is therefore expected to be

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 7 Vol.:(0123456789) www.nature.com/scientificreports/

transformed into hematite thermodynamically. During the initial stage of oxidation of Fe, outward difusion of Fe leads to the vacancy formation which later accumulates as cavities. Cavities are responsible for the formation of micro-channels that are known to pave the path for fast ­oxidation51. Corrosion flm is found with cracks which seem to be formed during the oxide scale growth on Agaria tribe’s-iron as illustrated by Nef et al. (2005)47. Tese cracks are expected to provide a path for continuous oxidation of Fe till the passive calcite and quartz phases are formed. It is noteworthy that precipitate phase formation depends on the localized presence of ions (e.g. ­Ca2+, 2- 47 52 ­CO3 , etc.) . However, the stability of these phases depends vastly on the ­atmosphere . Earlier studies have delineated the formation of crystalline iron hydrogen phosphate hydrate ­(FePO4·H3PO4·4H2O), α-, γ-, δ-FeOOH and magnetite in the case of Delhi iron pillar whereas Gupta dynasty’s Eran ­iron37,53 has shown the presence of γ -FeOOH and δ –FeOOH. It is worth mentioning that the research- ers were unable to detect phosphate phase (formed due to the phenomenon called micro-segregation) using Mössbauer spectroscopy which was identifed using micro-XRD54,55. In this study, the presence of hematite and maghemite were identifed as scale phases on the top of the iron made by the Agaria tribe. P is found present in slag whereas the presence of P in iron was not detected within the limit of the analytical techniques used in this study. On the basis of this result, we speculate application of lime and other basic compounds during the iron making process which would have led to the transfer P to slag. Te presence of slag elements such as Si, Ca (along with the Fe and O) in the scale grown on iron made by Agaria tribes is confrmed by XPS (Suppl. Figure 7). Oxides of these elements (Fe, Si and Ca) are generally known to be cathodic in nature and do not allow current to fow through and help in protecting iron from corrosion. Tese observations depict cathodic prevention as one of the reasons for the high corrosion resistance of the iron made through ancient Indian traditional methods. Second phase particles (such as slags and unreduced oxides) act as cathodic sites whereas metals such as Cr and Mn in the pure iron act as anodic sites. Second phase particles are known to form passive ­layers16. Te XPS survey scan does not show the presence of Al, which was noticed in FESEM-EDS (Suppl. Figure 4). Tis postulates that the Al-containing slag phase might have formed close to these layers and was beyond the resolution of the XPS instrument. Tis also means that the slag phases were formed at diferent depths within the corrosion scale formed on iron. It has been postulated earlier that mobile oxygen ions lead to the formation of a slag-oxide interface with the formation of large amounts of iron oxide products. In this study, iron corro- sion products along with the slag phases are detected in the flm formed on the top of the iron. We therefore speculate that atmospheric O played an important role in building up the inhibitive layers on the iron surface. It is observed from the FESEM-EDS mapping fgures (Fig. 5a,b) that Si, Ca, and Al—slag phase elements are not segregated but randomly distributed and can be therefore considered as mobile elements at the surface of the corrosion layer. Also, Fe was found depleted in a few instances. Similarly, these slag elements also remained non-depleted which indicates formation of microscale galvanic cells in the corrosion product layer. Tis highlights the slag-phases’ role in maintaining the cathodic passivity that provides corrosion protection to the underlying steel­ 56. We have also analysed the un-hammered and hammered iron with the help of neutron tomography (Fig. 6a,b) (see suppl. Figure 8 for the 3-D projection of diferent orientation views), which confrmed the presence of small volume pores. Hammering (forging) is known to remove the impurities (such as slag, pores, etc.) and to consolidate the pores available in materials. Te consolidation of pores in hammered iron can also be seen (Fig. 6b). Te values of the pores’ area (in 3-D) is found consolidated and most of them are found in the region of less than 1mm­ 2 in hammered iron, whereas un-hammered iron has exhibited the formation of large size pores, mostly found concentred within the region of 1 to ~ 3.5mm2 (see Suppl. Figure 9 (a)). Te majority of pores’ volume that is depicted in un-hammered and hammered iron are found below ~ 0.002 mm­ 3 and below ~ 0.04 ­mm3 respectively (see Suppl. Figure 9(b)). Te mathematical value of the pores’ volume in hammered iron is found higher than the un-hammered iron. On the other hand, the quantities of the pores inside the iron are also found higher in hammered iron. We speculated that this phenomenon could be attributed to the dislodgement of the inclusions (detachment of slag phases) from iron during the hammering (forging) operation. However, the superlative corrosion resistance of hammered iron could be associated with the thick passive corrosion product flm formation on the top of the iron as delineated earlier in this article which made iron remained inert. It is also noteworthy that the porosity obtained in neutron tomography was present inside the iron. A comparison of histograms of equi-diameters and length of the pores are shown in Suppl. Figure 10 (a) and (b). We also ana- lysed the pores present in iron made by Agaria tribes using synchrotron X-ray tomography. However, the low transmission and scattering through the sample material altered the quality of the data by introducing strong artefacts and prevented accurate porosities analysis. On the other hand, the neutron tomography images were free from these artefacts (See Suppl. Figure 11). In order to further relate the high corrosion resistance to the presence of a protective surface flm, we have carried out the analysis of the corrosion product flm thickness at the hot hammered (forged) and un-hammered (un-forged) iron. It is noticed that the thicker corrosion product flm was present on the hammered iron com- pared to the un-hammered iron. As illustrated earlier, metal-working process (mechanical metallurgical opera- tions) called hot hammering (forging) is known for removing impurities such as slags and consolidating internal pores, we hypothesize that the metal-working operation carried out by Agaria tribes has a great impact on making the iron corrosion resistance. We hypothesize that thick corrosion product layer, which is known to protect the substrate from further corrosion, formed on hammered iron and consolidated pores formation can play a vital role in iron’s corrosion resistance (See Fig. 7 (a) and (b); and Suppl. Figure 12).

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 8 Vol:.(1234567890) www.nature.com/scientificreports/

Figure 5. (a) FESEM images of corrosion product flm formed on the surface of iron made through ancient Indian metallurgical methods exhibits (b) EDS mapping of Fe, O, Si, P (dot points indicate the instrument noise), Ca, Al and, C.

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 9 Vol.:(0123456789) www.nature.com/scientificreports/

Figure 6. Neutron tomographic images depicting the volume of the pores that are formed on the iron made through ancient Indian metallurgical methods in (a) un-hammered iron (b) hammered (forged) iron.

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 10 Vol:.(1234567890) www.nature.com/scientificreports/

Figure 7. Neutron tomographic images of corrosion product flm thickness that are formed on the iron made through ancient Indian metallurgical methods in (a) un-hammered iron (b) hammered (forged) iron.

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 11 Vol.:(0123456789) www.nature.com/scientificreports/

Conclusions In summary, we have investigated the iron manufactured using the traditional iron making process followed by Indian tribes known as ‘Agaria’. Our results clearly demonstrated the importance of analytical techniques oper- ating at various length scale, in proposing the mechanisms of corrosion resistance including other properties of archaeomaterials. We have proposed the hypothesis for the excellent corrosion resistance of Agaria-tribe’s iron. In this article, the role of iron ore’s composition in the formation of corrosion products flm (protective flm against corrosion) on the iron surface is highlighted. Hematite and maghemite along with slag phases were depicted by using GIXRD and neutron difraction as the corrosion products on the iron surface. Te investiga- tions also highlighted the essential role of a multi-analytical techniques approach in the analysis of archaeoma- terials. Tese fndings were identifed as the primary reasons for the high corrosion resistance of the iron. Unlike Delhi iron pillar, P (which was found in the iron of Gupta’s period) is not found in the corrosion product layer formed on the iron (irrespective of hammered/un-hammered) manufactured by Agaria tribes within the limit of the correlative analytical techniques that are used in this study. However, the scope of further investigation still prevails. Te efect of metal-working operation (followed by Agaria tribes) in making the iron corrosion resistance is also demonstrated. We hypothesize that the metal working operation known as hot hammering, has led to the consolidation of internal pores (which were present in iron before pre-treatment) and removal of inclusions such as slag. Ticker passive corrosion product flm, which is known to be protective on the Agaria tribes’ iron, is observed for hammered iron than un-hammered iron. Hence, it is postulated that consolidation of pores, expected removal of inclusions and thick passive corrosion product layer formation have supported the iron in attaining the excellent corrosion resistance. We anticipate that our fndings will assist in uncovering the hidden knowledge pertaining to the archeomaterials’ (ferrous) degradation phenomenon, particularly in Japan and the South Asian region’s, with a history of close collaboration in ferrous metallurgy.

Received: 5 May 2020; Accepted: 17 December 2020

References 1. Singh, S. D. Iron in ancient India. J. Econ. Soc. Hist. Orient 5, 212–216 (1962). 2. Julef, G. An ancient wind-powered iron smelting technology in Sri Lanka. Nature 379, 60–63 (1996). 3. Bronson, B. Te making and selling of wootz, a crucible steel of India. Archeomaterials 1, 13–51 (1986). 4. Sukhanov, D. A., Arkhangel’skii, L. B. & Plotnikova, N. V. Nature of angular carbides in Damascus steel. Metallurgist 61, 40–46 (2017). 5. Park, J. S. & Shinde, V. Technology, chronology and the role of crucible steel as inferred from iron objects of the ancient site at Junnar India. J. Archaeol. Sci. 40, 3991–3998 (2013). 6. Park, J. S. & Shinde, V. Iron technology of the ancient megalithic communities in the Vidarbha region of India. J. Archaeol. Sci. 40, 3822–3833 (2013). 7. Reibold, M. et al. Carbon nanotubes in an ancient Damascus sabre. Nature 444, 286 (2006). 8. Possehl, G. L. & Gullapalli, P. Te early iron age in South Asia. in Te Archaeometallurgy of the Asian Old World (ed. Pigott, V. C.) 153–176 (University of Pennsylvania Museum, University Museum monograph, 1999). 9. Hudson, J. C. Te Delhi Pillar. Nature 172, 499–500 (1953). 10. Wranglén, G. Te, “rustless” iron pillar at Delhi. Corros. Sci. 10, 761–770 (1970). 11. Hadfeld, R. Discussion on Friend and Torneycraf’s paper on Ancient Iron. J. Iron Steel Res. Int. 112, 233–235 (1925). 12. Ghosh, M. K. Te Delhi iron pillar and its iron. NML Tech. J. 5, 31–45 (1963). 13. Balasubramaniam, R. Studies on the corrosion resistance of the Delhi iron pillar. NML Tech. J. 37, 123–142 (1995). 14. Balasubramaniam, R. Mixed potential theory analysis of the corrosion resistance of Delhi iron pillar. Trans. Indian Inst. Met. 50, 23–35 (1997). 15. Balasubramaniam, R., Kumar, A. V. R. & Dillmann, P. Characterization of rust on ancient Indian iron. Curr. Sci. 85, 1546–1555 (2003). 16. Balasubramaniam, R. On the corrosion resistance of the Delhi iron pillar. Corros. Sci. 42, 2103–2129 (2000). 17. Evans, U. R. Te Corrosion and Oxidation of Metals: Scientifc Principles and Practical Applications (Edward Arnold Publishers. LTD., London, 1960). 18. Soni, A. Iron-Metal Craf Economy of the Agaria Tribe of Chhattisgarh 370–375 (Aayu Publications, New Delhi, 2016). 19. Prakash, B. Ancient Indian iron and steel: An archaeometallurgical study. Indian J. Hist. Sci. 46, 381–410 (2011). 20. Igaki, K. High quality ancient wrought iron in Japan. in International Seminar on Wrought Iron, Society of Industrial Archaeology, Telford (1986). 21. Prakash, B. & Igaki, K. Ancient iron making in Bastar district. Indian J. Hist. Sci. 19, 172–185 (1984). 22. Dwivedi, D., Lepková, K. & Becker, T. Carbon steel corrosion: a review of key surface properties and characterization methods. RSC Adv. 7, 4580–4610 (2017). 23. Dwivedi, D., Lepkova, K. & Becker, T. Emerging surface characterization techniques for carbon steel corrosion: a critical brief review. Proc. Math. Phys. Eng. Sci. 473, 20160852 (2017). 24. Liss, K.-D., Hunter, B., Hagen, M., Noakes, T. & Kennedy, S. Echidna—the new high-resolution powder difractometer being built at OPAL. Phys. B Condens. Matter 385–386, 1010–1012 (2006). 25. Garbe, U. et al. A new neutron radiography/tomography/imaging station DINGO at OPAL. Phys. Procedia 69, 27–32 (2015). 26. Dierick, M., Masschaele, B. & Van Hoorebeke, L. Octopus, a fast and user-friendly tomographic reconstruction package developed in LabView. Meas. Sci. Technol. 15, 1366–1370 (2004). 27. Termo Scientifc™ Amira-Avizo Sofware. 28. Scarlett, N. V. Y., Tyson, P., Fraser, D., Mayo, S. & Maksimenko, A. Synchrotron X-ray CT characterization of titanium parts fab- ricated by additive manufacturing. Part I. Morphology. J. Synchrotron Radiat. 23, 1006–1014 (2016). 29. Wood, K., Mata, J. P., Garvey, C. J., Wu, C.-M., Hamilton, W. A., Abbeywick, P., Bartlett, D., Bartsch, F., Baxter, P., Booth, N., Brown, W., Christoforidis, J., Clowes, D., d’Adam, T., Darmann, F., Deura, M., Harrison, S., Hauser, N., Horton, G., Federici, D., Franceschini, F., Hanson, P., Imamovic, E., Imperia, P., Jones, M., Kennedy, S., Kim, S., Lam, T., Lee, W. T., Lesha, M., Mannicke, D., Noakes, T., Olsen, S. R., Osborn, J. C., Penny, D., Perry, M., Pullen, S. A., Robinson, R. A., Schulz, J. C., Xiong, N. & Gilbert, E. P. QUOKKA, the pinhole small-angle neutron scattering instrument at the OPAL Research Reactor, Australia: design, performance, operation and scientifc highlights. J. Appl. Cryst. 51, 294–314 (2018).

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 12 Vol:.(1234567890) www.nature.com/scientificreports/

30. Barone, G. et al. Small angle neutron scattering as fngerprinting of ancient potteries from Sicily (Southern Italy). J. Appl. Phys. 106, 054904 (2009). 31. Hammouda, B. A new Guinier-Porod model. J. Appl. Crystallogr. 43, 716–719 (2010). 32. Te available functional forms for SANS and USANS analysis, https​://www.ncnr.nist.gov/progr​ams/sans/data/avail​able_SANS. html. 33. Biswas, A. K. Iron and Steel in Pre-Modern India-A Critical Review. Indian J. Hist. Sci. 29, 579–610 (1994). 34. Alcántara, J., Chico, B., Simancas, J., Díaz, I. & Morcillo, M. Marine atmospheric corrosion of carbon steel: a review. Materials 10, 406 (2017). 35. Cox, A. & Lyon, S. B. An electrochemical study of the atmospheric corrosion of iron—II. Cathodic and anodic processes on uncor- roded and pre-corroded iron. Corros. Sci. 36, 1177–1192 (1994). 36. Stratmann, M., Bohnenkamp, K. & Engell, H. J. An electrochemical study of phase-transitions in rust layers. Corros. Sci. 23, 969–985 (1983). 37. Dillmann, P., Balasubramaniam, R. & Beranger, G. Characterization of protective rust on ancient Indian iron using microprobe analyses. Corros. Sci. 44, 2231–2242 (2002). 38. Moore, J. J. Chemical Metallurgy (2 edn) 152–192 (Butterworth-Heinemann, Oxford, 1990). 39. Elliott, R. Cast Iron Technology 91–125 (Butterworth-Heinemann, Oxford, 1988). 40. Devine, T. M. Corrosion of iron-base alloys. In Treatise on Materials Science & Technology (eds Briant, C. L. & Banerji, S. K.) 201–234 (Elsevier, Amsterdam, 1983). 41. Mitchell, A. Electrochemical Aspects of the ESR Process. IOP Conf. Ser. Mater. Sci. Eng. 143, 012001 (2016). 42. Majzlan, J., Grevel, K. D. & Navrotsky, A. Termodynamics of Fe oxides: Part II. Enthalpies of formation and relative stability of goethite (α-FeOOH), lepidocrocite (γ-FeOOH), and maghemite (γ-Fe2O3). Am. Mineral. 88, 855–859 (2003). 43. Hellmann, R. et al. Nanometre-scale evidence for interfacial dissolution–reprecipitation control of silicate glass corrosion. Nat. Mater. 14, 307–311 (2015). 44. Ackerman, K. J., Killick, D. J., Herbert, E. W. & Kriger, C. A. Study of Iron Smelting at Lopanzo, Equateur Province, Zaı̈re. J. Archaeol. Sci. 26, 1135–1143 (1999). 45. Kibanova, D., Trejo, M., Destaillats, H. & Cervini-Silva, J. Synthesis of hectorite–TiO2 and kaolinite–TiO2 nanocomposites with photocatalytic activity for the degradation of model air pollutants. Appl. Clay Sci. 42, 563–568 (2009). 46. Park, H., Kim, K. Y. & Choi, W. Photoelectrochemical approach for metal corrosion prevention using a semiconductor photoanode. J. Phys. Chem. B 106, 4775–4781 (2002). 47. Nef, D., Dillmann, P., Bellot-Gurlet, L. & Beranger, G. Corrosion of iron archaeological artefacts in soil: characterisation of the corrosion system. Corros. Sci. 47, 515–535 (2005). 48. Dwivedi, D. & Mata, J. P. Archaeometallurgical investigation of ancient artefacts’ degradation phenomenon. NPJ Mater. Degrad. 3, 1–12 (2019). 49. Navrotsky, A., Mazeina, L. & Majzlan, J. Size-driven structural and thermodynamic complexity in iron oxides. Science 319, 1635– 1638 (2008). 50. Birks, N., Meier, G., & Pettit, F. Mechanisms of oxidation. In Introduction to the High Temperature Oxidation of Metals. In Intro- duction to the High Temperature Oxidation of Metals 9–74 (Cambridge University Press, 2006). 51. Gibbs, G. B. & Hales, R. Te infuence of metal lattice vacancies on the oxidation of high temperature materials. Corros. Sci. 17, 487–507 (1997). 52. Ketteler, G., Weiss, W., Ranke, W. & Schlögl, R. Bulk and surface phases of iron oxides in an oxygen and water atmosphere at low pressure. Phys. Chem. Chem. Phys. 3, 1114–1122 (2001). 53. Kumar, A. V. R. & Balasubramaniam, R. Corrosion product analysis of corrosion resistant ancient Indian iron. Corros. Sci. 40, 1169–1178 (1998). 54. Balasubramaniam, R. Efect of material inhomogeneity on protective passive flm formation on Delhi iron pillar. Curr. Sci. 84, 534–541 (2003). 55. Balasubramaniam, R. Delhi iron pillar and its relevance to modern technology. Curr. Sci. 84, 126–127 (2003). 56. Koester, J. K. & Perkins, R. A. Discharge and corrosion characteristics of slagging metal electrodes for MHD power generators. J. Mater. Energy Syst. 1, 41–54 (1979). Acknowledgements We acknowledge technical assistance and facilities of the Curtin University Microscopy & Microanalysis Facility and the Australian Microscopy & Microanalysis Research Facility at the Centre for Microscopy, Characterisation & Analysis (CMCA), Te University of Western Australia (Prof. Martin Saunders), both of which are partially funded by the University, State, and Commonwealth Governments, the WA X-ray Surface Analysis Facility, funded by the Australian Research Council LIEF grant (LE120100026)). Te authors are grateful for the support from the School of Molecular and Life Sciences (Mr. Peter Chapman), Curtin University. Te authors thank Dr. Maxim Avdeev (Echidna, ANSTO) and Dr. Anton Maksimenko and Dr. Daniel Hausermann (IMBL, Austral- ian Synchrotron) for technical assistance with neutron difraction and synchrotron X-ray CT measurements, respectively. DD acknowledges Curtin University for a Curtin International Postgraduate Research Scholarship. DD thanks the Australian Institute of Nuclear Science and Engineering (AINSE) for their kind fnancial support under the AINSE Postgraduate Research Award (PGRA) scheme to enable this work. Authors acknowledge the Australian Nuclear Science and Technology Organization (ANSTO), Sydney (Quokka, Dingo (P6202) and Echidna (P6202) beamlines) and the Australian Synchrotron, Melbourne (IMBL beamline) (P 12293 and P 11768) for allowing us to use the respective beamlines. Te authors, particularly DD, greatly acknowledge the support and materials received from local Agaria tribes for this research. D. D. would like to dedicate this article to Professor S. P. Mehrotra (Visiting Professor, Indian Institute of Technology, Gandhinagar), and late Professor R. Balasubramaniam (Professor of Indian Institute of Technology, Kanpur), who was one of the pioneers in the feld of archaeometallurgy, and to his teachers. Author contributions D.D. collected the samples, proposed and designed the experiments. D.D. completed all experiments in this project, interpreted results, wrote the main manuscript and prepared supporting information materials including fgures. D.D. and M.R. performed XRD and analyzed difraction data. D.D. and F.S. performed neutron tomogra- phy and discussed the results. D.D. and J.M. performed SANS and discussed data. D.D., T.B. and K.L. discussed the results. D.D., T.B. and K.L. reviewed the fnal manuscript version. All authors participated in the discussion.

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 13 Vol.:(0123456789) www.nature.com/scientificreports/

Competing interests Te authors declare no competing interests. Additional information Supplementary Information Te online version contains supplementary material available at https​://doi. org/10.1038/s4159​8-021-81918​-w. Correspondence and requests for materials should be addressed to K.L. Reprints and permissions information is available at www.nature.com/reprints. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional afliations. Open Access Tis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. Te images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat​iveco​mmons​.org/licen​ses/by/4.0/.

© Te Author(s) 2021

Scientifc Reports | (2021) 11:4221 | https://doi.org/10.1038/s41598-021-81918-w 14 Vol:.(1234567890)